FPV LiPo Thermal Runaway Guard

Overview

An embedded-oriented battery-modeling study combines equivalent-circuit estimation, an extended Kalman filter, a reaction-rate proxy, and adaptive output limiting. Python processes public discharge data; C implements the estimation and control pipeline.

Problem definition

Voltage alone does not describe a battery’s state of charge, resistance growth, or thermal risk under changing load. The project studies how modeled state estimates can inform a response policy.

Research motivation

The research connects battery degradation estimation with high-current FPV use, while separating dataset comparisons and synthetic flight profiles from physical battery-abuse validation.

System architecture

A 1RC equivalent circuit feeds a five-state EKF with SOC, RC voltage, R0, R1, and C1. Resistance-based SOH is passed to an Arrhenius-style accumulated-reaction model.

controller.c applies NORMAL / WARNING / CRITICAL states with output limits of 100%, 70%, and 0%. The prototype tests these policies in software; it does not demonstrate a flight-qualified motor-control or fire-prevention system.

Methodology

Python derives OCV lookup tables and ECM parameters from NASA PCoE discharge records. C11 code performs estimation, a modeled remaining-time calculation, and policy updates. A separate RK4 FPV simulator supplies synthetic load and temperature profiles.

The reported comparisons cover 1RC versus 2RC circuits, coulomb counting versus EKF SOC, resistance- versus capacity-based SOH, temperature thresholds versus reaction-model alerts, and fixed versus adaptive output.

Experimental setup

The report uses NASA PCoE batteries B0005, B0006, B0007, and B0018, with a reported 168-cycle evaluation. A separate simulator repeats a hardcoded freestyle sequence.

The dataset’s slow sampling and the simulator’s simplified current and temperature models are part of the experimental conditions. Neither supplies measured FPV thermal-runaway onset times.

Results

The report describes approximately 22% lower SOC RMSE for EKF than coulomb counting, and 34% lower cumulative modeled risk-exposure time under adaptive limiting. These are the report’s comparison metrics, not independently verified safety outcomes.

The original comparison page below shows the resistance-based SOH, alert behavior, and policy comparison. Simulated time-to-critical over repeated endurance sequences must not be interpreted as physical flight endurance or a calibrated countdown to fire.

↗

Close ×

Discussion

The source implements the Arrhenius term with fixed reaction parameters and a degradation multiplier. That calculation is a model output whose physical calibration requires separate evidence.

The adaptive policy illustrates how estimates can affect system behavior. Cutting output in a simulation is distinct from establishing a safe response for an airborne aircraft.

Limitations

The report notes roughly 0.056 Hz dataset sampling and float-precision effects that can skip the WARNING state. Real-time STM32 deployment and actual FPV flight-data validation are further work.

NASA discharge-data comparison and synthetic flight simulation do not validate real thermal-runaway prevention. No physical prevention effectiveness or deployment qualification is claimed here.

References and resources

Sources: FPV-LiPo-Thermal-Runaway-Guard README, c/src/ekf.c, arrhenius.c, controller.c, and docs/Report_EN.pdf. The reviewed source revision and report are linked below.

Project resources

Content checked against the original report and linked implementation where available. Reported simulation or dataset results are distinct from physical validation.